Providers: OpenRouter, OpenAI, anything compatible

Codehamr speaks OpenAI Chat Completions and nothing else. Every provider that exposes that endpoint plugs in identically.

The whole guide is one rule and one table.

The rule

When a profile sets url:, codehamr POSTs to ${url}/v1/chat/completions. Always exactly that suffix.

So the url: value is the part before /v1/chat/completions:

Get this wrong and the first request 404s. Almost always the bug.

One config, every profile

active: local

models:
    local:
        llm: your-model
        url: http://localhost:11434
        key: ""
        context_size: 256000

    openrouter:
        llm: anthropic/claude-sonnet-4
        url: https://openrouter.ai/api
        key: sk-or-v1-...
        context_size: 128000

    openai:
        llm: gpt-5.5
        url: https://api.openai.com
        key: sk-proj-...
        context_size: 128000

    hamrpass:
        llm: hamrpass
        url: https://codehamr.com
        key: hp_...

Switch live inside a session with /models <name>.

Per-profile notes

• local, no key. llm: matches a tag from ollama list. Full setup in the Ollama guide. Codehamr in a container or WSL2 instead? Point url: at host.docker.internal (devcontainer) or use WSL mirrored mode, and start Ollama with OLLAMA_HOST=0.0.0.0 — see the devcontainer and WSL2 guides.
• openrouter. Any slug from openrouter.ai/models goes in llm:. Mirror the model's context length into context_size.
• openai. The API host is api.openai.com, not chatgpt.com. Different products.
• hamrpass. The llm: value is decorative, the server picks the model. No context_size needed: the server reports its own window via the X-Context-Window response header, and the budget bar in the TUI comes from X-Budget-Remaining.

One config, every provider, one switch.